German government plans to use AI in asylum procedures; experts warn of bias and errors
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
The German government has introduced a bill to deploy artificial intelligence in the country's asylum system, which will be debated in the Bundestag for the first time on Thursday evening. The legislation outlines three areas of AI use: analyzing completed asylum cases to identify patterns between applicant characteristics and decisions; using AI to verify dubious statements by applicants through biometric data and social media checks; and feeding data from past and current procedures into AI systems to train new models for future applications. While the government insists that final decisions will remain with human officials at the Federal Office for Migration and Refugees (Bamf), critics warn of significant risks. Experts and opposition politicians, including Greens and Left party members, argue that AI systems may reproduce existing biases from training data, leading to more prejudiced outcomes. They also highlight the danger of automation bias, where officials may over-rely on AI recommendations, and the risk of AI 'hallucinations' producing false information. The Left party's Clara Bünger accused the government of treating refugees as 'guinea pigs' for a technology not yet applied to German citizens. The bill has been criticized as a 'blank check' for AI deployment without sufficient transparency or safeguards.
Source report
Experts warn that even if AI programs only assist caseworkers, the risk of erroneous decisions remains high.
The German federal government plans to introduce artificial intelligence into the asylum system. A corresponding draft law is set to be debated in the Bundestag for the first time on Thursday evening. While the government insists that final decisions will remain in human hands, the Greens, the Left Party, and numerous experts have expressed alarm over the proposals, which aim to significantly accelerate asylum procedures.
Three Areas of AI Deployment
The draft law outlines three specific applications for AI programs in asylum proceedings:
- Systematic evaluation of completed cases – AI will analyze past asylum decisions to identify correlations between applicants' personal characteristics and outcomes. For example, certain names may indicate membership in a persecuted minority or origin from a crisis region. If individuals with such names have historically been granted asylum more frequently, the AI would recognize this pattern.
- Informing future case assessments – Insights derived from these patterns will be made available to decision-makers at the Federal Office for Migration and Refugees (Bamf). This is intended to influence "the future intensity of scrutiny in individual case decisions." In practice, this could mean that applicants with certain characteristics receive either superficial or particularly thorough examination.
- Verification of applicant statements – Bamf staff will be able to use AI to cross-check dubious claims made by asylum seekers. Biometric data—primarily photographs—will be automatically compared with information found online, including social media profiles.
Additionally, the draft law permits feeding data from both completed and ongoing proceedings into AI systems to train new models and explore further applications. Critics note that this indicates a vague intention to expand AI use beyond what is currently planned.
Risks of Bias and Error
Although the government emphasizes that human officials will retain final decision-making authority, experts warn that providing staff with AI-generated data opens multiple pathways to error.
Studies show that people tend to rely heavily on automated assessments, particularly under stress. If Bamf personnel effectively rubber-stamp AI recommendations, the question arises as to where true decision-making power actually lies.
Moreover, AI programs do not produce objective results. They replicate the biases present in their training data. If previous asylum decisions were influenced by racist prejudices against a particular social group, the AI will reproduce those tendencies. The result could be even more one-sided decisions that disproportionately harm affected individuals.
The use of AI to verify applicant statements is also fraught with risk. Even the latest AI chat programs continue to generate false information—so-called "hallucinations." Conversely, if Bamf staff were to manually verify every piece of AI-provided information, the hoped-for efficiency gains would likely evaporate.
Political and Expert Reactions
Clara Bünger, interior policy spokesperson for the Left Party, stated: "Refugees are being used as guinea pigs for a technology that no one would dare apply to German citizens with such intensity." She noted that the draft law is scheduled for debate late in the evening, when few parliamentarians or journalists are present. "One could not express more clearly how little regard there is for the fundamental rights of those affected."
Lukas Benner, the Greens' spokesperson on the interior committee, commented: "In principle, there is nothing wrong with using AI in public administration." However, he stressed that clear rules and transparency are essential. The government's draft law, he said, is the opposite—"a blank check for the deployment of arbitrary AI systems."
Independent experts have also voiced serious concerns. The German Informatics Society (Gesellschaft für Informatik) cited "significant reservations regarding technical feasibility, compatibility with data protection, and the guarantee of rule-of-law oversight mechanisms."
Source
taz.de - taz.deWestern